git.sesse.net Git - mlt/blob - src/modules/sox/filter_sox.c

   1 /*
   2  * filter_sox.c -- apply any number of SOX effects using libst
   3  * Copyright (C) 2003-2004 Ushodaya Enterprises Limited
   4  * Author: Dan Dennedy <dan@dennedy.org>
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  19  */
  20
  21 #include <framework/mlt_filter.h>
  22 #include <framework/mlt_frame.h>
  23 #include <framework/mlt_tokeniser.h>
  24 #include <framework/mlt_log.h>
  25
  26 #include <stdio.h>
  27 #include <stdlib.h>
  28 #include <string.h>
  29 #include <math.h>
  30
  31 // TODO: does not support multiple effects with SoX v14.1.0+
  32
  33 #ifdef SOX14
  34 #       include <sox.h>
  35 #       define ST_EOF SOX_EOF
  36 #       define ST_SUCCESS SOX_SUCCESS
  37 #       define st_sample_t sox_sample_t
  38 #       define eff_t sox_effect_t*
  39 #       define ST_LIB_VERSION_CODE SOX_LIB_VERSION_CODE
  40 #       define ST_LIB_VERSION SOX_LIB_VERSION
  41 #       if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,2,0))
  42 #               define st_size_t size_t
  43 #       else
  44 #               define st_size_t sox_size_t
  45 #       endif
  46 #       define ST_SIGNED_WORD_TO_SAMPLE(d,clips) SOX_SIGNED_16BIT_TO_SAMPLE(d,clips)
  47 #       if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
  48 #               define ST_SSIZE_MIN SOX_SAMPLE_MIN
  49 #       else
  50 #               define ST_SSIZE_MIN SOX_SSIZE_MIN
  51 #       endif
  52 #               define ST_SAMPLE_TO_SIGNED_WORD(d,clips) SOX_SAMPLE_TO_SIGNED_16BIT(d,clips)
  53 #else
  54 #       include <st.h>
  55 #endif
  56
  57 #define BUFFER_LEN 8192
  58 #define AMPLITUDE_NORM 0.2511886431509580 /* -12dBFS */
  59 #define AMPLITUDE_MIN 0.00001
  60
  61 /** Compute the mean of a set of doubles skipping unset values flagged as -1
  62 */
  63 static inline double mean( double *buf, int count )
  64 {
  65         double mean = 0;
  66         int i;
  67         int j = 0;
  68
  69         for ( i = 0; i < count; i++ )
  70         {
  71                 if ( buf[ i ] != -1.0 )
  72                 {
  73                         mean += buf[ i ];
  74                         j ++;
  75                 }
  76         }
  77         if ( j > 0 )
  78                 mean /= j;
  79
  80         return mean;
  81 }
  82
  83 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
  84 static void delete_effect( eff_t effp )
  85 {
  86         free( effp->priv );
  87         free( (void*)effp->in_encoding );
  88         free( effp );
  89 }
  90 #endif
  91
  92 /** Create an effect state instance for a channels
  93 */
  94 static int create_effect( mlt_filter this, char *value, int count, int channel, int frequency )
  95 {
  96         mlt_tokeniser tokeniser = mlt_tokeniser_init();
  97         char id[ 256 ];
  98         int error = 1;
  99
 100         // Tokenise the effect specification
 101         mlt_tokeniser_parse_new( tokeniser, value, " " );
 102         if ( tokeniser->count < 1 )
 103                 return error;
 104
 105         // Locate the effect
 106         mlt_destructor effect_destructor = mlt_pool_release;
 107 #ifdef SOX14
 108         //fprintf(stderr, "%s: effect %s count %d\n", __FUNCTION__, tokeniser->tokens[0], tokeniser->count );
 109 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 110         sox_effect_handler_t const *eff_handle = sox_find_effect( tokeniser->tokens[0] );
 111         if (eff_handle == NULL ) return error;
 112         eff_t eff = sox_create_effect( eff_handle );
 113         effect_destructor = ( mlt_destructor ) delete_effect;
 114         sox_encodinginfo_t *enc = calloc( 1, sizeof( sox_encodinginfo_t ) );
 115         enc->encoding = SOX_ENCODING_SIGN2;
 116         enc->bits_per_sample = 16;
 117         eff->in_encoding = eff->out_encoding = enc;
 118 #else
 119         eff_t eff = mlt_pool_alloc( sizeof( sox_effect_t ) );
 120         sox_create_effect( eff, sox_find_effect( tokeniser->tokens[0] ) );
 121 #endif
 122         int opt_count = tokeniser->count - 1;
 123 #else
 124         eff_t eff = mlt_pool_alloc( sizeof( struct st_effect ) );
 125         int opt_count = st_geteffect_opt( eff, tokeniser->count, tokeniser->tokens );
 126 #endif
 127
 128         // If valid effect
 129         if ( opt_count != ST_EOF )
 130         {
 131                 // Supply the effect parameters
 132 #ifdef SOX14
 133 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,2,0))
 134                 if ( sox_effect_options( eff, opt_count, &tokeniser->tokens[ tokeniser->count > 1 ? 1 : 0  ] ) == ST_SUCCESS )
 135 #else
 136                 if ( ( * eff->handler.getopts )( eff, opt_count, &tokeniser->tokens[ tokeniser->count > 1 ? 1 : 0  ] ) == ST_SUCCESS )
 137 #endif
 138 #else
 139                 if ( ( * eff->h->getopts )( eff, opt_count, &tokeniser->tokens[ tokeniser->count - opt_count ] ) == ST_SUCCESS )
 140 #endif
 141                 {
 142                         // Set the sox signal parameters
 143 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 144                         eff->in_signal.rate = frequency;
 145                         eff->out_signal.rate = frequency;
 146                         eff->in_signal.channels = 1;
 147                         eff->out_signal.channels = 1;
 148                         eff->in_signal.precision = 16;
 149                         eff->out_signal.precision = 16;
 150                         eff->in_signal.length = 0;
 151                         eff->out_signal.length = 0;
 152 #else
 153                         eff->ininfo.rate = frequency;
 154                         eff->outinfo.rate = frequency;
 155                         eff->ininfo.channels = 1;
 156                         eff->outinfo.channels = 1;
 157 #endif
 158
 159                         // Start the effect
 160 #ifdef SOX14
 161                         if ( ( * eff->handler.start )( eff ) == ST_SUCCESS )
 162 #else
 163                         if ( ( * eff->h->start )( eff ) == ST_SUCCESS )
 164 #endif
 165                         {
 166                                 // Construct id
 167                                 sprintf( id, "_effect_%d_%d", count, channel );
 168
 169                                 // Save the effect state
 170                                 mlt_properties_set_data( MLT_FILTER_PROPERTIES( this ), id, eff, 0, effect_destructor, NULL );
 171                                 error = 0;
 172                         }
 173                 }
 174         }
 175         // Some error occurred so delete the temp effect state
 176         if ( error == 1 )
 177                 effect_destructor( eff );
 178
 179         mlt_tokeniser_close( tokeniser );
 180
 181         return error;
 182 }
 183
 184 /** Get the audio.
 185 */
 186
 187 static int filter_get_audio( mlt_frame frame, void **buffer, mlt_audio_format *format, int *frequency, int *channels, int *samples )
 188 {
 189 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,3,0))
 190         SOX_SAMPLE_LOCALS;
 191 #endif
 192         // Get the filter service
 193         mlt_filter filter = mlt_frame_pop_audio( frame );
 194
 195         // Get the filter properties
 196         mlt_properties filter_properties = MLT_FILTER_PROPERTIES( filter );
 197
 198         mlt_service_lock( MLT_FILTER_SERVICE( filter ) );
 199
 200         // Get the properties
 201         st_sample_t *input_buffer;// = mlt_properties_get_data( filter_properties, "input_buffer", NULL );
 202         st_sample_t *output_buffer = mlt_properties_get_data( filter_properties, "output_buffer", NULL );
 203         int i; // channel
 204         int count = mlt_properties_get_int( filter_properties, "_effect_count" );
 205         int analysis = mlt_properties_get( filter_properties, "effect" ) && !strcmp( mlt_properties_get( filter_properties, "effect" ), "analysis" );
 206
 207         // Get the producer's audio
 208         *format = mlt_audio_s32;
 209         mlt_frame_get_audio( frame, buffer, format, frequency, channels, samples );
 210
 211         // Even though some effects are multi-channel aware, it is not reliable
 212         // We must maintain a separate effect state for each channel
 213         for ( i = 0; i < *channels; i++ )
 214         {
 215                 char id[ 256 ];
 216                 sprintf( id, "_effect_0_%d", i );
 217
 218                 // Get an existing effect state
 219                 eff_t e = mlt_properties_get_data( filter_properties, id, NULL );
 220
 221                 // Validate the existing effect state
 222 #if (ST_LIB_VERSION_CODE >= ST_LIB_VERSION(14,1,0))
 223                 if ( e != NULL && ( e->in_signal.rate != *frequency ||
 224                                                         e->out_signal.rate != *frequency ) )
 225 #else
 226                 if ( e != NULL && ( e->ininfo.rate != *frequency ||
 227                                                         e->outinfo.rate != *frequency ) )
 228 #endif
 229                         e = NULL;
 230
 231                 // (Re)Create the effect state
 232                 if ( e == NULL )
 233                 {
 234                         int j = 0;
 235
 236                         // Reset the count
 237                         count = 0;
 238
 239                         // Loop over all properties
 240                         for ( j = 0; j < mlt_properties_count( filter_properties ); j ++ )
 241                         {
 242                                 // Get the name of this property
 243                                 char *name = mlt_properties_get_name( filter_properties, j );
 244
 245                                 // If the name does not contain a . and matches effect
 246                                 if ( !strncmp( name, "effect", 6 ) )
 247                                 {
 248                                         // Get the effect specification
 249                                         char *value = mlt_properties_get_value( filter_properties, j );
 250
 251                                         // Create an instance
 252                                         if ( create_effect( filter, value, count, i, *frequency ) == 0 )
 253                                                 count ++;
 254                                 }
 255                         }
 256
 257                         // Save the number of filters
 258                         mlt_properties_set_int( filter_properties, "_effect_count", count );
 259
 260                 }
 261                 if ( *samples > 0 && ( count > 0 || analysis ) )
 262                 {
 263                         input_buffer = (st_sample_t*) *buffer + i * *samples;
 264                         st_sample_t *p = input_buffer;
 265                         st_size_t isamp = *samples;
 266                         st_size_t osamp = *samples;
 267                         int j = *samples + 1;
 268                         char *normalise = mlt_properties_get( filter_properties, "normalise" );
 269                         double normalised_gain = 1.0;
 270
 271                         if ( analysis )
 272                         {
 273                                 // Run analysis to compute a gain level to normalize the audio across entire filter duration
 274                                 double max_power = mlt_properties_get_double( filter_properties, "_max_power" );
 275                                 double peak = mlt_properties_get_double( filter_properties, "_max_peak" );
 276                                 double use_peak = mlt_properties_get_int( filter_properties, "use_peak" );
 277                                 double power = 0;
 278                                 int n = *samples + 1;
 279
 280                                 // Compute power level of samples in this channel of this frame
 281                                 while ( --n )
 282                                 {
 283                                         double s = fabs( *p++ );
 284                                         // Track peak
 285                                         if ( s > peak )
 286                                         {
 287                                                 peak = s;
 288                                                 mlt_properties_set_double( filter_properties, "_max_peak", peak );
 289                                         }
 290                                         power += s * s;
 291                                 }
 292                                 power /= *samples;
 293                                 // Track maximum power
 294                                 if ( power > max_power )
 295                                 {
 296                                         max_power = power;
 297                                         mlt_properties_set_double( filter_properties, "_max_power", max_power );
 298                                 }
 299
 300                                 // Complete analysis the last channel of the last frame.
 301                                 if ( i + 1 == *channels && mlt_filter_get_position( filter, frame ) + 1
 302                                          == mlt_filter_get_length2( filter, frame ) )
 303                                 {
 304                                         double rms = sqrt( max_power / ST_SSIZE_MIN / ST_SSIZE_MIN );
 305                                         char effect[32];
 306
 307                                         // Convert RMS or peak to gain
 308                                         if ( use_peak )
 309                                                 normalised_gain = ST_SSIZE_MIN / -peak;
 310                                         else
 311                                                 normalised_gain = AMPLITUDE_NORM / rms;
 312
 313                                         // Set properties for serialization
 314                                         snprintf( effect, sizeof(effect), "vol %f", normalised_gain );
 315                                         effect[31] = 0;
 316                                         mlt_properties_set( filter_properties, "effect", effect );
 317                                         mlt_properties_set( filter_properties, "analyze", NULL );
 318
 319                                         // Show output comparable to normalize --no-adjust --fractions
 320                                         mlt_properties_set_double( filter_properties, "level", rms );
 321                                         mlt_properties_set_double( filter_properties, "gain", normalised_gain );
 322                                         mlt_properties_set_double( filter_properties, "peak", -peak / ST_SSIZE_MIN );
 323                                 }
 324
 325                                 // restore some variables
 326                                 p = input_buffer;
 327                         }
 328
 329                         if ( normalise )
 330                         {
 331                                 int window = mlt_properties_get_int( filter_properties, "window" );
 332                                 double *smooth_buffer = mlt_properties_get_data( filter_properties, "smooth_buffer", NULL );
 333                                 double max_gain = mlt_properties_get_double( filter_properties, "max_gain" );
 334                                 double rms = 0;
 335
 336                                 // Default the maximum gain factor to 20dBFS
 337                                 if ( max_gain == 0 )
 338                                         max_gain = 10.0;
 339
 340                                 // Compute rms amplitude
 341                                 while( --j )
 342                                 {
 343                                         rms += ( double )*p * ( double )*p;
 344                                         p ++;
 345                                 }
 346                                 rms = sqrt( rms / *samples / ST_SSIZE_MIN / ST_SSIZE_MIN );
 347
 348                                 // The smoothing buffer prevents radical shifts in the gain level
 349                                 if ( window > 0 && smooth_buffer != NULL )
 350                                 {
 351                                         int smooth_index = mlt_properties_get_int( filter_properties, "_smooth_index" );
 352                                         smooth_buffer[ smooth_index ] = rms;
 353
 354                                         // Ignore very small values that adversely affect the mean
 355                                         if ( rms > AMPLITUDE_MIN )
 356                                                 mlt_properties_set_int( filter_properties, "_smooth_index", ( smooth_index + 1 ) % window );
 357
 358                                         // Smoothing is really just a mean over the past N values
 359                                         normalised_gain = AMPLITUDE_NORM / mean( smooth_buffer, window );
 360                                 }
 361                                 else if ( rms > 0 )
 362                                 {
 363                                         // Determine gain to apply as current amplitude
 364                                         normalised_gain = AMPLITUDE_NORM / rms;
 365                                 }
 366
 367                                 //printf("filter_sox: rms %.3f gain %.3f\n", rms, normalised_gain );
 368
 369                                 // Govern the maximum gain
 370                                 if ( normalised_gain > max_gain )
 371                                         normalised_gain = max_gain;
 372                         }
 373
 374                         // For each effect
 375                         for ( j = 0; j < count; j++ )
 376                         {
 377                                 sprintf( id, "_effect_%d_%d", j, i );
 378                                 e = mlt_properties_get_data( filter_properties, id, NULL );
 379
 380                                 // We better have this guy
 381                                 if ( e != NULL )
 382                                 {
 383                                         float saved_gain = 1.0;
 384
 385                                         // XXX: hack to apply the normalised gain level to the vol effect
 386 #ifdef SOX14
 387                                         if ( normalise && strcmp( e->handler.name, "vol" ) == 0 )
 388 #else
 389                                         if ( normalise && strcmp( e->name, "vol" ) == 0 )
 390 #endif
 391                                         {
 392                                                 float *f = ( float * )( e->priv );
 393                                                 saved_gain = *f;
 394                                                 *f = saved_gain * normalised_gain;
 395                                         }
 396
 397                                         // Apply the effect
 398 #ifdef SOX14
 399                                         if ( ( * e->handler.flow )( e, input_buffer, output_buffer, &isamp, &osamp ) != ST_SUCCESS )
 400 #else
 401                                         if ( ( * e->h->flow )( e, input_buffer, output_buffer, &isamp, &osamp ) != ST_SUCCESS )
 402 #endif
 403                                         {
 404                                                 mlt_log_warning( MLT_FILTER_SERVICE(filter), "effect processing failed\n" );
 405                                         }
 406
 407                                         // XXX: hack to restore the original vol gain to prevent accumulation
 408 #ifdef SOX14
 409                                         if ( normalise && strcmp( e->handler.name, "vol" ) == 0 )
 410 #else
 411                                         if ( normalise && strcmp( e->name, "vol" ) == 0 )
 412 #endif
 413                                         {
 414                                                 float *f = ( float * )( e->priv );
 415                                                 *f = saved_gain;
 416                                         }
 417                                 }
 418                         }
 419
 420                         // Write back
 421                         memcpy( input_buffer, output_buffer, *samples * sizeof(st_sample_t) );
 422                 }
 423         }
 424
 425         mlt_service_unlock( MLT_FILTER_SERVICE( filter ) );
 426
 427         return 0;
 428 }
 429
 430 /** Filter processing.
 431 */
 432
 433 static mlt_frame filter_process( mlt_filter this, mlt_frame frame )
 434 {
 435         if ( mlt_frame_is_test_audio( frame ) == 0 )
 436         {
 437                 // Add the filter to the frame
 438                 mlt_frame_push_audio( frame, this );
 439                 mlt_frame_push_audio( frame, filter_get_audio );
 440
 441                 // Parse the window property and allocate smoothing buffer if needed
 442                 mlt_properties properties = MLT_FILTER_PROPERTIES( this );
 443                 int window = mlt_properties_get_int( properties, "window" );
 444                 if ( mlt_properties_get( properties, "smooth_buffer" ) == NULL && window > 1 )
 445                 {
 446                         // Create a smoothing buffer for the calculated "max power" of frame of audio used in normalisation
 447                         double *smooth_buffer = (double*) calloc( window, sizeof( double ) );
 448                         int i;
 449                         for ( i = 0; i < window; i++ )
 450                                 smooth_buffer[ i ] = -1.0;
 451                         mlt_properties_set_data( properties, "smooth_buffer", smooth_buffer, 0, free, NULL );
 452                 }
 453         }
 454
 455         return frame;
 456 }
 457
 458 /** Constructor for the filter.
 459 */
 460
 461 mlt_filter filter_sox_init( mlt_profile profile, mlt_service_type type, const char *id, char *arg )
 462 {
 463         mlt_filter this = mlt_filter_new( );
 464         if ( this != NULL )
 465         {
 466                 void *input_buffer = mlt_pool_alloc( BUFFER_LEN );
 467                 void *output_buffer = mlt_pool_alloc( BUFFER_LEN );
 468                 mlt_properties properties = MLT_FILTER_PROPERTIES( this );
 469
 470                 this->process = filter_process;
 471
 472                 if ( !strncmp( id, "sox.", 4 ) )
 473                 {
 474                         char *s = malloc( strlen( id ) + ( arg? strlen( arg ) + 2 : 1 ) );
 475                         strcpy( s, id + 4 );
 476                         if ( arg )
 477                         {
 478                                 strcat( s, " " );
 479                                 strcat( s, arg );
 480                         }
 481                         mlt_properties_set( properties, "effect", s );
 482                         free( s );
 483                 }
 484                 else if ( arg )
 485                         mlt_properties_set( properties, "effect", arg );
 486                 mlt_properties_set_data( properties, "input_buffer", input_buffer, BUFFER_LEN, mlt_pool_release, NULL );
 487                 mlt_properties_set_data( properties, "output_buffer", output_buffer, BUFFER_LEN, mlt_pool_release, NULL );
 488                 mlt_properties_set_int( properties, "window", 75 );
 489                 mlt_properties_set( properties, "version", sox_version() );
 490         }
 491         return this;
 492 }
 493
 494 // What to do when a libst internal failure occurs
 495 void cleanup(void){}